Benzodiazepine (BZD) use in the U.S. is common and increases with age. In a recent analysis, 8.7% of adults aged 65-80 years were prescribed BZDs during one year, even though a robust set of studies have established their association with a variety of adverse outcomes in older adults, including increased risk of falls and fractures, motor vehicle accidents, impaired cognition, and pharmaceutical overdose. Patients and their providers are then reluctant to change use once started, which may account for why older adults experience the highest rates of long-term BZD use. Relatively little is known about the patient, provider, and community characteristics associated with starting and continuing BZD prescribing to older adults, yet this is critical to develop effective selective and indicated prevention strategies. In Aim 1, we will describe the patient, provider, and community characteristics associated with BZD initiation and continuation using a national 20% sample of Medicare beneficiaries (n=3.6 million) linked to provider data from the American Medical Association (AMA) Physician Masterfile and community characteristics from the Area Health Resources File (AHRF). We will extend our analysis with OptumInsight data (n=6.7 million) to gain additional insights among commercially insured adults aged 50-64 given increased substance use among the Baby Boom cohort. Those patients currently prescribed BZDs and most at risk for BZD misuse (e.g., overlapping BZD prescriptions from multiple providers) and BZD-related overdose should receive indicated prevention strategies to address this potentially harmful use. In Aim 2, among those prescribed BZD, we will determine specific risk factors associated with BZD misuse and BZD-related overdose; these data will be used to develop a misuse clinical prediction tool. Using BZD users 50+ years old identified in Medicare and Optum, we will determine characteristics of patients and their prescribed BZD (e.g., high potency) most associated with misuse and overdose. We will then use machine learning to create a simple clinical prediction tool that providers can use to identify older adults at risk for misuse in their practices. Finally, in Aim 3 we will conduct semi-structured interviews with providers and patients to package and script the use of the clinical prediction tool for providers seeking to engage high-risk BZD use patients. This aim is critical to improve the impact of our findings since psychological dependence on BZD can make reducing use a difficult topic for physicians and patients to address. We will conduct interviews with providers and older adult primary care patients (n=15 each) to obtain feedback to package and script the use of the clinical prediction tool, which we will make publicly available by website. The impact of our work will be to: 1) provide a detailed, national portrait of the factors that contribute to BZD use and misuse; 2) determine the older adults most at risk for serious adverse events; and 3) develop and package a clinical prediction tool to help providers address BZD use in their high-risk patients.